 |
Center for Nonlinear Studies |
Both clustering and biclustering are widely-used techniques in data mining, and have found many applications in pattern recognition, image analysis, information retrieval, text mining, bioinformatics, etc. In this talk, we use graph partitioning approaches for clustering and biclustering. For different proposes of clustering and biclustering, three kinds of cuts for graph partitioning, including minimum cut, ratio cut and normalized cut, are formulated by mathematical programming approaches. Many optimization methods, such as integer programming, spectral and semidefinite programming relaxations, are proposed for solving real problems. Numerical experiments based on randomly generated data sets are performed for comparing these methods.
Host: Sasha Gutfraind, T-5, CNLS, gutfraind@lanl.gov